1. Single-channel currents evoked by high concentrations of GABA (10-2000 microM) have been analysed to investigate the characteristics of GABAA receptor channels in outside-out patches from rat sympathetic neurones. When high concentrations of GABA were applied to a patch, channel openings occurred in prolonged clusters (3.8 +/- 3.7 s (mean +/- S.D.) at 50 microM-GABA) consisting, on average, of 350 apparent openings per cluster. Individual clusters were separated by long silent intervals. 2. Channel openings were to many (often ill-defined) conductance states (range 7-36 pS), but the most frequently observed conductance level was approximately 30 pS, (29.6 +/- 0.34 pS). Only these clusters during which the channel was open to this main state conductance for at least 95% of the cluster open time were used in the analysis of probability of being open. Other less frequently observed conductance levels were 15-18 and 22-23 pS, while levels of 33-36 and 7-9 pS were occasionally, but reliably, observed. 3. Bursts within clusters were defined as a series of openings separated by closed intervals shorter than some critical value, tc. At 50 microM-GABA the mean burst length was 439 +/- 434 ms (+/- S.D., tc = 50 ms). 4. The probability of being open, po, during bursts within clusters has been analysed as a function of GABA concentration. As expected, increasing the concentration of GABA resulted in an overall increase in po. However, for a given agonist concentration there was a wide spread in po, far greater than that predicted for a population of identical and independent receptor channels (demonstrated by comparison with stimulated channel activity). 5. The wide range of po values at a particular concentration of GABA was not due to inappropriate selection of tc. At 50 microM-GABA the range of po values was similar for tc of 20-1000 ms, although the overall mean po became lower (0.64 rather than 0.81). 6. On the basis of simulated channel activity, it appears that most clusters which do not contain multiple openings, arise from the activity of one individual channel. Furthermore, there was no detectable tendency for gaps between bursts to be shorter in the middle of a cluster than at its ends. Therefore it is unlikely that variability in po arose from overlapping activity of two or more channels. 7. The values for po, mean open time, and mean shut time for bursts within the same cluster, and between different clusters, were compared by a randomization test.(ABSTRACT TRUNCATED AT 400 WORDS)